The use of automatic learning algorithms that include socioeconomic data could be key to predicting edentulism.
The scientific journal PLOS ONE, which belongs to the US Public Library of Science, published last June the study “Predictors of tooth loss: A machine learning approach”, which highlights the importance of Artificial Intelligence (AI) as a tool capable of predicting tooth loss.
The study was conducted based on the medical and socioeconomic data of nearly 12,000 American adults from the US National Health and Nutrition Examination Survey (NHANES) (2011-2014), a key factor in the effectiveness of the algorithms used in predicting tooth loss and functional dentition.
As Dr Hawazin Elani, assistant professor of oral health policy and epidemiology at Harvard School of Dental Medicine and the lead author of the study, explains: “Our analysis showed that while all machine learning models can be useful for predicting risk, those that incorporate socioeconomic variables can be particularly powerful screening tools for identifying people at higher risk of tooth loss”.
The study incorporated the socio-economic factors of patients’ age, education, employment, income, race or ethnicity into the analysis, which resulted to be more important predictors of tooth loss than the patients’ medical factors.
The results show, how important socio-economic levels can be for population health. As Elani points out: “Our results suggest that machine learning algorithm models that incorporate socioeconomic characteristics are better at predicting tooth loss than those based solely on routine clinical dental indicators.
The medical data analysed included diseases such as arthritis, diabetes, high cholesterol, hypertension and heart disease, which were also predictive of tooth loss but not with the results obtained with the socio-economic factors.
For the study, five different machine learning algorithms were analysed and developed to predict complete and incremental tooth loss. For each of them, their predictive performance was evaluated by examining the area under the receiver operating characteristics curve (AUC), which measures parameters of accuracy, sensitivity, specificity as well as positive results and negative predictive values.
The AUC is a metric used to evaluate the performance of algorithms. The closer the AUC is to 100%, the better the algorithm is at predicting between classes, in this case, tooth loss or no tooth loss.
The results of the study showed that all algorithms performed well with an AUC of over 86.5% on average. The highest result was for edentulism which achieved an AUC of 89%. Functional dentition and missing teeth also performed very well with an AUC of 88% and 83% respectively.
